Abstract:
|
'Statistical Challenges in Assessing and Fostering the Reproducibility of Scientific Results'(NAM press, 2016) stresses that sharing of i) data, ii) statistical code, and iii) dynamic reporting (DR) are essential for ensuring reproducibility of research (RR). In the field of oncology, RR defined as the ability of others to reproduce the published findings, given the original data, has been problematic. We hypothesize that readiness of statisticians with 1) knowledge of tools for making a collaborative project easy to share and 2) leadership skills for engaging collaborators with discussion about the issue of non-RR are essential for improvement. We took 3 steps at a cancer center biostatistics core: 1) trained statisticians on data archiving and dynamic report generation (R, R Markdown, Knitr); 2) developed illustrative examples of studies with longitudinally collected quality of life endpoints and with assessment of prognostic value of gene signature; both needed repeated data updates and benefitted from DR; 3) organized a 'Rigor and Reproducibility Journal Club" for discussion, collaboration, and training. We will discuss emerging successes and barriers.
|